On the Nonparametric Approach to Estimation of Non- Constant Variance Function: An Application to Nairobi Securities Exchange (NSE)

Methods for estimating regression models to data in the areas showing varying variances is considered. The centre of attention is on diverse methods of evaluating varying variances. The nonparametric approach which incorporates the smoothing methods and the choice of the ideal bandwidth is discussed. Normally, the cardinal shortcoming which is of interest is the selection of the smoothing method and picking of the best bandwidth [1], Zhai, C. and Lafferty, J. [2]. The two oftenly used smoothing methods; the Gaussian Kernel and Spline are compared. The two smoothing techniques are illustrated and compared using data obtained from Nairobi Securities exchange (NSE) and found that the Gaussian Kernel outperforms the Spline smoother since it gives the best estimate of the variance.